EP3791361B1 - Tumoral mass detection system based on magnetic resonance imaging - Google Patents

Tumoral mass detection system based on magnetic resonance imaging Download PDF

Info

Publication number
EP3791361B1
EP3791361B1 EP19730217.7A EP19730217A EP3791361B1 EP 3791361 B1 EP3791361 B1 EP 3791361B1 EP 19730217 A EP19730217 A EP 19730217A EP 3791361 B1 EP3791361 B1 EP 3791361B1
Authority
EP
European Patent Office
Prior art keywords
values
parameters
voxel
voxels
cell
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
EP19730217.7A
Other languages
German (de)
English (en)
French (fr)
Other versions
EP3791361A1 (en
Inventor
Valentina Giannini
Simone MAZZETTI
Daniele REGGE
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fondazione Del Piemonte Per L'oncologia
Universita degli Studi di Torino
Original Assignee
Fondazione Del Piemonte Per L'oncologia
Universita degli Studi di Torino
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Fondazione Del Piemonte Per L'oncologia, Universita degli Studi di Torino filed Critical Fondazione Del Piemonte Per L'oncologia
Publication of EP3791361A1 publication Critical patent/EP3791361A1/en
Application granted granted Critical
Publication of EP3791361B1 publication Critical patent/EP3791361B1/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/143—Segmentation; Edge detection involving probabilistic approaches, e.g. Markov random field [MRF] modelling
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves 
    • A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/285—Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G06V20/698—Matching; Classification
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2576/00—Medical imaging apparatus involving image processing or analysis

Definitions

  • the present invention concerns a tumoral mass detection system based on magnetic resonance imaging (MRI).
  • MRI magnetic resonance imaging
  • the diagnosis of tumoral masses in particular in the case of tumoral masses in the prostate, has various criticalities such as, for example, low sensitivity and specificity of the screening and diagnosis methods. These criticalities negatively affect the patient quality of life and cause an increase in costs for the national health system.
  • PSA Prostatic Specific Antigen
  • the so-called PSA test is carried out, which evaluates blood PSA levels in the patient under examination. If the PSA values are high, the patient undergoes a prostate biopsy.
  • the PSA test has a low specificity and low sensitivity ( ⁇ 30%); this means that approximately 60% of patients with high PSA values, but who do not present significant tumoral masses, are subjected to unnecessary prostate biopsy, undergoing the side effects connected with this operation. On the other hand, patients with potentially serious tumours do not undergo prostate biopsy due to the low levels of PSA detected.
  • the prostate biopsy performed by means of random bioptic sampling, is not always able to provide a complete representation of the tumoral masses. It is known, in fact, that the concordance between the aggressiveness of the tumour at the time of the prostate biopsy and post-biopsy varies between 28% and 69%.
  • tumours can be evaluated as non-aggressive and, consequently, can be under-treated; similarly, patients with non-aggressive tumours can be subjected to radical treatments (i.e. they can be over-treated) when, on the contrary, they could benefit from less invasive treatments with fewer side effects.
  • the current MRI systems extract a certain number of parameters, starting from MRI images previously acquired, and subsequently determine the presence or absence of tumoral masses on the basis of said parameters.
  • the Applicant has observed that, although the MRI systems represent an undoubted step forward with respect to previous diagnostic techniques, the diagnostic precision provided by them is still subject to improvement.
  • the object of the present invention is therefore to provide a detection system based on MRI images which allows improvement in the precision of the diagnosis of tumoral masses.
  • a detection system for detecting tumoral masses is provided, as defined in the attached claims.
  • Figure 1 shows a system 1, which comprises an MRI apparatus 2, of a known type, and a processing system 4, which are electrically coupled to each other.
  • system 1 is an image processing system and a diagnostic aid, of computer-assisted type, therefore it is referred to below as CAD system 1.
  • CAD system 1 is described with reference to the article " A fully automatic computer aided diagnosis system for peripheral zone prostate cancer detection using multi-parametric magnetic resonance imaging” (Giannini V., Mazzetti S., Regge D. et al., Computerized Medical Imaging and Graphics 2015; 46:219-226 ), highlighting the differences of the present CAD system 1.
  • the CAD system 1 operates on a selected population of sample patients (for example, including Q sample patients), who have previously undergone radical prostatectomy; in this way, it is possible to know a priori the bioptic result relative to the presence of a prostate tumour, so as to allow a sort of preliminary calibration of the CAD system 1.
  • the MRI apparatus 2 acquires (block 10) a plurality of MRI type images, using different image acquisition techniques.
  • the MRI apparatus 2 uses MRI acquisition techniques to perform scans based on the following image types:
  • the MRI apparatus 2 includes known instruments such as, for example, a scanner using a first coil, having four or more channels in phase (four-channel phased-array), in combination with a second coil, of endorectal type and arranged in the vicinity of the prostate.
  • known instruments such as, for example, a scanner using a first coil, having four or more channels in phase (four-channel phased-array), in combination with a second coil, of endorectal type and arranged in the vicinity of the prostate.
  • figure 3 shows, in a simplified manner, how any one of the previously mentioned scans is performed.
  • figure 3 schematically shows a prostate 15 of any sample patient, having a base 15A and an apex 15B, crossed by a sagittal axis S, which ideally connects the base 15A to the apex 15B; furthermore, the sagittal axis S is parallel to a first axis Z of a Cartesian reference system XYZ and is perpendicular to planes parallel to a plane XY of the same Cartesian reference system XYZ.
  • the MRI apparatus 2 scans the prostate 15, so as to generate a plurality of axial slices (two shown in figure 3 ), indicated overall by the reference notation F i ; in detail, i is an integer index with value between 1 and N, where N is the number of axial slices Fi used to perform the section scan of the prostate 15. Furthermore, each axial slice Fi is parallel to the plane XY of the Cartesian reference system XYZ and is, therefore, perpendicular to the sagittal axis S of the prostate 15.
  • each axial slice Fi the MRI apparatus 2 generates a corresponding image; in turn, each image is formed by a corresponding plurality of minimum image units, defined as voxels.
  • each voxel is representative of a corresponding slice portion of prostatic tissue; in this regard, the above-mentioned slices have thickness, for example, of three millimetres.
  • the processing system 4 receives and stores the images generated by the MRI apparatus 2, if necessary, after storing the images in a database of the MRI apparatus 2. Furthermore, according to the MRI acquisition technique considered, the images of the corresponding scan are processed in a different manner by the processing system 4.
  • the T2w acquisition technique allows N morphological images of the prostate 15 to be acquired relative to the same instant of time t*, so as to allow observation of the variations of the first relaxation time T2 between the axial slices F i .
  • the DWI acquisition technique allows N morphological images of the prostate 15 to be acquired, relative to the same instant of time t**. More precisely, the DWI acquisition technique acquires at least two corresponding groups of N images each so that, in a first approximation, both are relative to the above-mentioned instant of time t** and allow determination, based on said at least two groups, of N images relative to the above-mentioned coefficient ADC.
  • the DWI N images reference to the N images relative to the ADC parameter is understood, unless specified otherwise.
  • the DCE-MRI acquisition technique it is possible to acquire N images of the prostate 15 for each of k instants of time, equally spaced from one another.
  • the acquisition is performed starting from an initial instant of time to' and terminates at a final instant of time t' k-1 ; furthermore, in each instant of time of the time interval t 0 '-t' k-1 N functional images of the prostate 15 are acquired.
  • a period for example of thirteen seconds, elapses.
  • the DCE-MRI acquisition technique allows spatial and temporal acquisition of the images, so that at the end of the acquisition process, we have k ⁇ N images of the prostate 15. Furthermore, starting from the acquired images, it is possible to trace, for each voxel, a corresponding time evolution curve, which is defined as contrast uptake curve; in detail, the contrast uptake curve is formed by k corresponding points and is indicative of the time evolution of the distribution of the contrast medium in the corresponding tissue portion of the prostate 15.
  • the DCE-MRI acquisition technique associates each voxel with one or more parameters indicative of the corresponding contrast uptake curve.
  • the parameters indicative of the contrast uptake curve may be semi-quantitative or quantitative.
  • the semi-quantitative parameters are calculated starting from the values of the contrast uptake curve and are, for example, maximum absorption (maximum uptake, MU), peak time (time to peak, TTP), the wash-in rate and the washout rate.
  • the quantitative parameters are obtained by interpolating, for each voxel, the corresponding contrast uptake curve according to two categories of interpolation models; in the case presented, the interpolation occurs, for example, according to the Tofts pharmacokinetic model and by means of some mathematical models described below.
  • the Tofts pharmacokinetic model allows the determination of specific microvascular parameters of the tumour, such as capillary permeability, blood flow and blood volume; these parameters are calculated by interpolating in a known manner each contrast uptake curve relative to each voxel.
  • a corresponding voxel three-dimensional matrix 45 (an example of which is shown in figure 4 ) is stored, for example in the processing system 4.
  • each voxel is associated with one or more parameters (features) relative to the corresponding portion of prostatic tissue, which are stored by the processing system 4.
  • each voxel of the corresponding matrix 45 is associated with a corresponding MRI signal intensity value.
  • each voxel of the corresponding matrix 45 is associated with a corresponding MRI signal intensity value.
  • each voxel is associated with the values of the above-mentioned one or more semi-quantitative and quantitative parameters obtained from the corresponding contrast uptake curve.
  • the value/values of a voxel are referred to below to indicate, considering a given acquisition technique, the value/values of the corresponding parameter/s associated with the voxel in question.
  • the corresponding matrix 45 has a cubic form and is formed of voxels also having cubic form.
  • the matrix 45 does not depend on the acquisition technique; in fact, even if the three acquisition techniques involved, for example, voxels of different dimensions, it would be possible to achieve the described scenario by means of known processing techniques (for example, interpolation).
  • the voxels are indicated overall by the notation V(a, b, c), in which "a”, “b” and “c” are indexes which represent the spatial position of each voxel with respect to the Cartesian reference system XYZ.
  • "a” is the index representative of the spatial position along a second axis X of the Cartesian reference system XYZ
  • "b” is the index representative of the spatial position along a third axis Y of the Cartesian reference system XYZ
  • “c” is the index representative of the spatial position along the first axis Z.
  • the index c indicates the axial slice F i to which the single voxel V(a, b, c) belongs.
  • the voxels V(a, b, c) are represented so that voxels of successive images are arranged in contact with one another, although in reality voxels of successive images may refer to portions of prostate not necessarily adjacent to one another. In any case, below, the vertically aligned voxels of pairs of successive images are called adjacent, even when the corresponding portions of prostate are not adjacent.
  • index c is an integer variable between 1 and N
  • the processing system 4 performs an alignment operation (block 20) of the images acquired, independently of the MRI acquisition technique used to obtain said images.
  • the images acquired by means of the above-mentioned three acquisition techniques refer to a same portion of prostate; in other words, voxels that have the same position in the three matrixes 45 acquired by means of the above-mentioned three MRI acquisition techniques refer to a same sub-portion of prostate.
  • the scans performed based on the above-mentioned image acquisition techniques refer to the same voxels. For this reason, below we will refer generically to a voxel to indicate the corresponding sub-portion of prostate, unless specified otherwise.
  • each voxel is associated with a corresponding group of parameter values, referred to below as initial parameters.
  • the parameter values include: at least one value equal to the value of said voxel in the image acquired with the T2w acquisition technique; at least one value equal to the value of the parameter ADC of said voxel; and one or more values equal to the one or more values of the semi-quantitative and quantitative parameters obtained from the corresponding contrast uptake curve.
  • the processing system 4 performs (block 30) an automatic segmentation step of the prostate 15.
  • This automatic segmentation allows discarding of the areas external to the prostate 15 which may have been scanned previously.
  • the segmentation does not modify the images previously acquired.
  • further operations may be performed that allow reduction of the dimensions of the images acquired; however, for the sake of simplicity it is assumed that also such possible further operations do not modify the images acquired.
  • the processing system 4 calculates (block 40), for each voxel, at least one additional parameter, referred to below as the structural parameter (texture).
  • the processing system 4 selects one of the above-mentioned initial parameters. Subsequently, the processing system 4 calculates the value of the structural parameter relative to the voxel 51, based on:
  • AP a b c P a b c + f P a ⁇ l , b ⁇ m , c ⁇ o
  • f indicates a generic dependence
  • 1, m and o are binary variables such that the binary string "lmo" assumes at least one sub-group of the seven values 001, 010, 011, 100, 101, 110 and 111.
  • the one or more additional parameters are each indicative of the spatial distribution of the corresponding initial parameter, in a neighbourhood of the voxel considered.
  • the definition of said neighbourhood may vary with respect to what is described; even more generally, the definition of neighbourhood may vary between the different structural parameters.
  • each voxel V(a, b, c) is associated with a corresponding plurality of extracted parameters, understood as including both the initial parameters and the structural parameters.
  • the total number of extracted parameters is equal to J.
  • a number equal to J of corresponding parameter values are stored, which refer to the prostate 15 of any one sample patient.
  • the processing system 4 performs a selection step (block 60) of a sub-group of the J parameters extracted, which are referred to below as significant parameters.
  • an operator associates with each voxel of the sample patient a binary indication relative to the presence/absence of tumour in said voxel, based on the pathological indications provided by the prostate samples. Said association is also stored in the processing system 4.
  • the processing system 4 stores (block 61, figure 6 ), for each sample patient, a flat matrix 47 and a vector 48, shown for example in figure 7 .
  • the flat matrix 47 has N ⁇ M 2 lines which correspond to the voxels. Furthermore, the flat matrix 47 shows, for each voxel, the corresponding values of the extracted parameters. On the other hand, the vector 48 contains, for each voxel, the corresponding binary indication (indicated as Ps, while the extracted parameters are indicated by P p , where p is an integer index ranging from 1 to J). In the example of figure 7 , the flat matrix 47 and the vector 48 refer, for example, to a first sample patient.
  • the processing system 4 determines (block 62, figure 6 ), for each of the extracted parameters, a corresponding ROC (Receiver Operating Characteristic) curve, which indicates the ability of each parameter to distinguish between malignant tissue and healthy tissue, and calculates the area below (also defined as AUROC, Area Under ROC).
  • ROC Receiveiver Operating Characteristic
  • the processing system 4 selects the columns (in a number equal to Q) of the flat matrixes 47 relative to the sample patients corresponding to the extracted parameter considered, obtaining a macrocolumn 300 (shown in figure 8 and referring, for example and without any loss of generality, to the first extracted parameter P 1 ), given by the succession of the columns selected. Furthermore, said macrocolumn 300 is associated with a macrovector 310 (shown in figure 8 ) given by the succession of the vectors 48 relative to the sample patients.
  • the processing system 4 arranges the values of the macrocolumn 300 in increasing order; this operation entails a corresponding reordering of the binary indications of the macrovector 310, so as to maintain the original voxel-binary indication associations.
  • the processing system 4 repeats the following operations, for each of the values of the extracted parameter considered (below indicated by W):
  • the processing system 4 is therefore able to store, for each threshold value previously set, a corresponding point of the ROC curve referring to the corresponding extracted parameter considered; in particular, as shown in figure 9 , said point has an X axis equal to the difference between one and the specificity 1-Sp(W) and a Y axis equal to the sensitivity S(W). Subsequently, the processing system 4 interpolates the points and obtains the corresponding ROC curve; for example, as mentioned previously, figure 9 shows an ROC 1 curve that refers to the first extracted parameter P 1 .
  • the processing system 4 has as many ROC curves as columns of each flat matrix 47; in particular, in the case considered, the processing system 4 generates J ROC curves.
  • the processing system 4 calculates the area below each ROC curve, therefore associating a numeric value with each extracted parameter P p .
  • the processing system 4 determines (block 63), therefore, a correlation matrix, by means of the linear correlation method in pairs.
  • An example of correlation matrix is shown in figure 10 and is indicated by the reference number 320; in particular, this correlation matrix 320 is constructed so that the extracted parameters are organised, on the lines and on the columns, in decreasing order (from top to bottom, and therefore also from left to right) based on the area below the respective ROC curve.
  • the greater the area below the ROC curve the greater the specificity and the sensitivity associated with the corresponding parameter extracted for determining the presence/absence of a tumoral mass.
  • a corresponding correlation coefficient is defined, based on the corresponding pair of macrocolumns 300.
  • the correlation coefficient varies between -1 (in the case of inversely correlated parameters) and +1 (in the case of directly correlated parameters) and is equal to zero in the case of totally uncorrelated parameters.
  • the correlation matrix 320 is a symmetrical matrix J ⁇ J , on the diagonal of which the correlation of each parameter with itself (therefore, equal to 1) is shown and in the other positions the correlation coefficients of the pairs of extracted parameters are shown.
  • each correlation parameter is indicated by the reference ⁇ e,f , in which e and f are indexes referring to the extracted parameters, the calculation of the correlation of which is desired.
  • the processing system 4 selects (block 64, figure 6 ) the significant parameters, based on the correlation matrix 320 constructed in the preceding step, in which the parameters were ordered in decreasing values of the area below the ROC curve.
  • the processing system 4 analyses in sequence all the correlation coefficients of the over-diagonal or under-diagonal half of the correlation matrix 320. For example, assuming that the over-diagonal half of the correlation matrix 320 is analysed, it is analysed by lines, from top to bottom, and from left to right. Furthermore, when a pair of extracted parameters has a degree of correlation equal to or greater than 80% (i.e.
  • the processing system 4 chooses the parameter of the pair whose ROC curve subtends the larger area, and discards the other; this means that, proceeding in the analysis of the correlation coefficients, the correlation coefficients that involve the discarded parameter will not be considered. The analysis therefore continues without considering the parameters that are gradually discarded, as far as the last correlation coefficient of the over-diagonal half. In practice, whenever a pair of extracted parameters with correlation coefficient equal to or greater than 80% meet, only the parameter most representative of the presence/absence of prostate tumoral masses is chosen (maintained), therefore the processing system 4 continues to consider it as a possible significant parameter, unlike the discarded parameter.
  • the processing system 4 selects (block 65 of figure 2 ), within the flat matrix 47 referring to each sample patient, only the columns relative to the significant parameters, thus reducing the number of columns of the flat matrix 47 relative to each sample patient.
  • a reduced matrix 49 is obtained, an example of which is shown in figure 11 (in which the significant parameters are indicated by the same signs as the corresponding extracted parameters, with the addition of a superscript).
  • the reduced matrix 49 shown in figure 11 refers, for example, to the first sample patient and has N ⁇ M 2 lines, representing the voxels, and S columns, representing the significant parameters.
  • the processing system 4 determines (block 70) a first classifier, based on the values of the significant parameters P p ' contained in the reduced matrixes 49 of the sample patients and the corresponding binary indications indicative of the presence of tumour.
  • the processing system 4 generates and stores a classifier of known type (for example, SVM, Support Vector Machine classifier, as described for example in " Machine learning in medical imaging", IEEE Signal Process Magazine, 2010 July, 27(4): 25-38, di M.N. Wernick et al. ) , i.e. a mathematical model adapted to receive as input data the values of the significant parameters P p ' of any one voxel of any one patient and generate at output a corresponding value indicative of the probability that said voxel represents a tumoral voxel.
  • a classifier of known type for example, SVM, Support Vector Machine classifier, as described for example in " Machine learning in medical imaging", IEEE Signal Process Magazine, 2010 July, 27(4): 25-38, di M.N. Wernick et al.
  • a mathematical model adapted to receive as input data the values of the significant parameters P p ' of any one voxel of any one patient and generate at output a corresponding value indicative of the probability that said vo
  • the group of the probability values forms a corresponding map, which associates, with each voxel V(a, b, c), a corresponding probability of representing a tumoral voxel; this map is referred to below as three-dimensional likelihood map.
  • the processing system 4 carries out a selection (block 80) of the voxels of the corresponding three-dimensional likelihood map which have probability values greater than or equal to, for example, 60%.
  • the processing system 4 stores, for each planar matrix of voxels, the areas formed by voxels which respond to the selection criterion described (probability value greater than or equal to, for example, 60%) and are furthermore adjacent to one another; therefore, the above-mentioned areas represent corresponding isolated selected voxels or corresponding aggregates of selected voxels. Below this areas will be referred to as two-dimensional regions of voxels.
  • figure 12 refers to the first sample patient and shows a section 200 of the corresponding three-dimensional likelihood map, in which a plurality of two-dimensional regions of voxels R 1_2D -R 5_2D and a portion of remaining map 202 are present.
  • the portion of remaining map 202 is the portion of the section 200 complementary to the plurality of two-dimensional regions of voxels R 1_2D -R 5_2D ; the portion of remaining map 202 therefore represents the group of voxels of the section 200 having probability of presenting a tumoral mass (for example) lower than 60%.
  • the processing system 4 determines (block 90) the dimension of each of the two-dimensional regions of voxels, which is equal to the number of voxels constituting each of the above-mentioned two-dimensional regions of voxels.
  • the processing system 4 compares (block 100) the dimension of each two-dimensional region of voxels with a predefined threshold value, for example equal to 100mm 2 . In particular, if the dimensions of a two-dimensional region of voxels are lower than the threshold value, the processing system 4 discards the region (block 110). Below, we refer to the two-dimensional regions selected to indicate the two-dimensional regions of voxels having dimensions greater than the threshold value.
  • processing system 4 carries out a further selection step (optional), in which, in each two-dimensional region selected, any "noisy" voxels are removed.
  • a voxel is defined "noisy" if considered as representative of a false positive.
  • the processing system 4 analyses the contrast uptake curve referring to each voxel forming the selected two-dimensional region under investigation.
  • a voxel is indicative of the actual presence of a prostate tumour if, after approximately 60 seconds from injection of the contrast medium (for example, at the time instant t 4 ' or at the time instant ts', in the case of temporal resolution equal to 13 seconds), the contrast uptake curve presents a rapid ascent; consequently, it is possible to discard the voxels having less marked contrast uptake curves without the rapid ascent described, since these voxels are not effectively representative of prostate tissues affected by tumour.
  • the step of removal of the noisy voxels is not carried out.
  • the processing system 4 carries out a step of determination (block 120) of three-dimensional regions of voxels.
  • a three-dimensional region of voxels is defined when at least two selected two-dimensional regions, belonging to two planar matrixes of adjacent voxels, are connected, or are at least partially overlapped along the first axis Z of the Cartesian reference system XYZ.
  • figure 13 shows a first and a second planar matrix of voxels (indicated by the references F g , F g+1 of the corresponding axial slices, in which g is an index between 1 and N-1), each including its own plurality of selected two-dimensional regions.
  • the first planar matrix of voxels F g comprises a first and a second selected two-dimensional region R 1_2D , R 2_2D ;
  • the second planar matrix of voxels F g+1 comprises a third, a fourth, a fifth and a sixth selected two-dimensional region R 1_2D ' , R 2_2D ' , R 3_2D ' and R 4_2D '.
  • the first selected two-dimensional region R 1_2D is vertically aligned with the third selected two-dimensional region R 1_2D '; analogously, the second selected two-dimensional region R 2_2D is vertically aligned with the fourth selected two-dimensional region R 2_2D '.
  • the fifth and the sixth selected two-dimensional regions R 3_2D ', R 4_2D ' overlap the remaining map portion 252 of the planar matrix of voxels F g .
  • the first and the second selected two-dimensional regions R 1_2D , R 2_2D are respectively connected to the third and the fourth selected two-dimensional regions R 1_2D ', R 2_2D '.
  • the first and the second selected two-dimensional regions of voxels R 1_2D , R 2_2D have the same forms as the third and the fourth selected two-dimensional regions of voxels R 1_2D ' , R 2_2D '.
  • the processing system 4 stores the three-dimensional regions of voxels previously determined.
  • the processing system 4 calculates, for each three-dimensional region of voxels, at least one regional parameter (block 130).
  • the processing system 4 selects one of the initial parameters. Subsequently, for each three-dimensional region of voxels, the processing system 4 calculates the corresponding value of the regional parameter based on the values of the initial selected parameter of the voxels that form the three-dimensional region of voxels.
  • possible regional parameters can be given by parameters indicative of the contrast and/or homogeneity of the three-dimensional regions of voxels, calculated on the basis of the values of corresponding initial parameters.
  • the processing system 4 may calculate values indicative, respectively: of the homogeneity of the intensity of the MRI signal acquired with the T2w acquisition technique; of the contrast of the intensity of the MRI signal acquired with the T2w acquisition technique; of the homogeneity of the ADC parameter acquired with the DWI acquisition technique; of the contrast of the ADC parameter acquired with the DWI acquisition technique; of the homogeneity of the intensity of the MRI signal acquired with the DCE acquisition technique in any one of the above-mentioned instants t 0 '-t' k-1 ; and of the contrast of the intensity of the MRI signal acquired with the acquisition technique DCE in any one of the above-mentioned instants t0'-t'k-1.
  • Other values may also be calculated such as, for example, entropy values, as described in R. M. Haralick, K. Shanmugam, and I. Dinstein, "Textural Features of Image Classification", IEEE Transactions on Systems, Man and Cybernetics, vol. SMC-3, no. 6, Nov. 1973 , or energy values, again based on the signals acquired with one or more of the above-mentioned acquisition techniques.
  • entropy values as described in R. M. Haralick, K. Shanmugam, and I. Dinstein, "Textural Features of Image Classification", IEEE Transactions on Systems, Man and Cybernetics, vol. SMC-3, no. 6, Nov. 1973
  • energy values again based on the signals acquired with one or more of the above-mentioned acquisition techniques.
  • Further examples of possible regional parameters are represented by statistical parameters (for example, the mean, median and percentiles) of the intensity of the MRI signal acquired according to one or more of the three acquisition techniques.
  • the processing system 4 stores, for each sample patient, the regional parameters determined in the preceding step (block 140).
  • the regional parameters determined in the preceding step block 140.
  • a number equal to L of values of corresponding regional parameters is stored.
  • the processing system 4 carries out a selection step (block 150) of a sub-group of the L regional parameters, referred to below as significant regional parameters.
  • An operator associates with each three-dimensional region of voxels of each sample patient a binary indication relative to the degree of aggressiveness of the tumour in said three-dimensional region of voxels.
  • each three-dimensional region of voxels considered undergoes a preliminary evaluation by a pathologist.
  • each three-dimensional region of voxels is associated by the pathologist with two numerical references according to the known Gleason score system.
  • each three-dimensional region of voxels is assigned:
  • the Gleason score I 3 D is obtained, defined as a combination of in _1 and in _2 .
  • the Gleason score I 3 D can be equal to 4+3.
  • the operator then associates '0' with the three-dimensional regions of voxels having Gleason score I 3 D lower than or equal to (for example) 3+3 and associates '1' with the three-dimensional regions of voxels having Gleason score I 3 D higher than 3+3.
  • the processing system 4 stores (block 151), for each sample patient, a regional flat matrix 247 and a regional vector 248, shown for example in figure 16 , relative to the first sample patient.
  • the regional flat matrix 247 has as many lines as the number of three-dimensional regions of voxels identified for the corresponding sample patient; in the example shown in figure 16 , the first sample patient has V three-dimensional regions of voxels.
  • the regional flat matrix 247 shows, for each three-dimensional region of voxels, the corresponding values of the regional parameters; therefore, the regional flat matrix 247 has dimensions equal to V ⁇ L.
  • the regional vector 248 contains, for each three-dimensional region of voxels, the corresponding binary indication relative to the aggressiveness (indicated as P S T , while the parameters are indicated by P r T , where r is an integer index ranging from 1 to L).
  • the processing system 4 determines (block 152), for each of the regional parameters, a corresponding ROC T regional curve and the relative area, in the same way as discussed with reference to block 62, with the exception of the fact that the calculation is performed based on a regional macrocolumn 400 ( figure 17 ) given by the columns (in a number equal to Q) of the regional flat matrixes 247 which correspond to the regional parameter considered, and based on a regional macrovector 410 given by the succession of regional vectors 248 relative to the sample patients.
  • the processing system 4 determines (block 153) a regional correlation matrix (shown in figure 18 , where it is indicated by 420), carrying out the same operations described with reference to block 63, with the exception of the fact that these operations are performed on the regional macrocolumns 400.
  • the processing system 4 selects (block 154) the significant regional parameters, based on the regional correlation matrix 420, performing the same operations as those described for block 64, without prejudice to the possibility of adopting a different threshold value from the one used in the operations in block 64.
  • the processing system 4 selects (block 160), from within the regional flat matrix 247 of each sample patient, only the columns relative to the significant regional parameters. In this way, a corresponding reduced regional matrix 249 is obtained, an example of which, relative to the first sample patient, is shown in figure 19 , in which the significant regional parameters are indicated by the addition of a superscript.
  • the processing system 4 determines (block 170 of figure 2 ) a second classifier (for example, of Bayes type, or decision tree, SVM (as described for example in Kononenko, I., I. Bratko and M. Kukar, 1998. Application of Machine Learning to Medical Diagnosis. In: Machine Learning and Data Mining: Methods and Applications, R.S. Michalski, I. Bratko and M. Kubat (Eds.). J. Wiley, New York ), based on the values of the significant regional parameters P r T' contained in the reduced regional matrixes 249 of the sample patients and of the corresponding binary indications P S T .
  • SVM Bayes type, or decision tree
  • the processing system 4 generates and stores a mathematical model adapted to receive as input data the values of the significant regional parameters P r T' of any one three-dimensional region of voxels of any one patient and generate at output a corresponding indication of the degree of aggressiveness of the tumoral mass present in said three-dimensional region of voxels.
  • the processing system 4 selects (block 171) the three-dimensional regions of voxels associated with '1', which are referred to below as relevant three-dimensional regions.
  • the processing system 4 divides (block 172) the relevant three-dimensional region into a plurality of cells, as shown for example in figure 20 , where the cells (four) are indicated by 373 and the relevant three-dimensional region is indicated by R 3D *.
  • the cells 373 have for example the same form (predefined), so as to form a regular lattice.
  • the cells 373 are each formed of a corresponding group of adjacent voxels (for example, in the shape of parallelepipeds or cubes); furthermore, the cells 373 are also adjacent to one another, so as to cover, as a whole, the entire relevant three-dimensional region. In practice, each cell covers a corresponding portion of the relevant three-dimensional region to which it belongs.
  • the processing system 4 calculates (block 173) a corresponding group of parameter values, which are referred to below as cell parameters.
  • the cell parameters may be respectively equal to the above-mentioned regional parameters, apart from the fact that they are calculated on the domain of the cell, instead of on the entire three-dimensional region of voxels.
  • the cell parameters may therefore be formed, for example, from parameters indicative of the contrast and/or the homogeneity and/or the entropy and/or the energy and/or of statistical parameters of the cells of voxels, calculated on the basis of the values of corresponding initial parameters of the voxels that form the cells.
  • the number of cell parameters is, for example, equal to D.
  • the processing system 4 stores (block 174), for each sample patient, the cell parameters determined in the preceding step.
  • the processing system 4 carries out a new selection step (block 175) of a sub-group of the cell parameters, which will be referred to below as significant cell parameters.
  • An operator associates with each cell of each sample patient a binary indication relative to the degree of aggressiveness of the tumour in said cell.
  • each cell undergoes a preliminary evaluation by a pathologist, who associates a value '1' if he/she considers the cell aggressive (for example, if it has a Gleason score higher than or equal to three), or associates a value '0' if he/she considers the cell non-aggressive.
  • the processing system 4 stores (block 180), for each sample patient, a flat cell matrix 347 and a cell vector 348, shown for example in figure 22 , relative to the first sample patient.
  • the flat cell matrix 347 has as many lines as the cells identified for the corresponding sample patient; in the example shown in figure 22 , the first sample patient has overall ⁇ cells, independently of how these cells are shared between the relevant three-dimensional regions of the first sample patient.
  • the flat cell matrix 347 shows, for each cell, the corresponding values of the cell parameters; therefore, the flat cell matrix 347 has dimensions equal to ⁇ D .
  • the cell vector 348 contains, for each cell, the corresponding binary indication relative to the aggressiveness (indicated as P S T* , while the cell parameters are indicated by P r T* ).
  • the processing system 4 determines (block 181), for each of the cell parameters, a corresponding ROC T* cell curve and the relative area, in the same way as discussed with reference to block 62, without prejudice to the fact that the calculation is performed on the basis of a cell macrocolumn 450 ( figure 23 ) given by the columns of the flat cell matrixes 347 which correspond to the cell parameter considered, and on the basis of a cell macrovector 460 given by the succession of the cell vectors 348 relative to the sample patients.
  • the processing system 4 determines (block 182) a cell correlation matrix (shown in figure 24 , where it is indicated by 470), performing the same operations as those described for block 63, without prejudice to the fact that these operations are performed on the cell macrocolumns 450. Subsequently, the processing system 4 selects (block 183) the significant cell parameters, based on the cell correlation matrix 470, performing the same operations as those described for block 64, without prejudice to the possibility of adopting a different threshold value from the one used in the operations in block 64.
  • a cell correlation matrix shown in figure 24 , where it is indicated by 470
  • the processing system 4 selects (block 176), from within the flat cell matrix 347 of each sample patient, only the columns relative to the significant cell parameters. In this way, a corresponding reduced cell matrix 349 is obtained, an example of which, relative to the first sample patient, is shown in figure 25 , in which the significant cell parameters are indicated by the addition of a superscript.
  • the processing system 4 determines (block 177 of figure 2 ) a third classifier (for example, of Bayes type, or decision tree or SVM), based on the values of the significant cell parameters P r T' * contained in the reduced cell matrixes 349 of the sample patients and of the corresponding binary indications P S T *.
  • a third classifier for example, of Bayes type, or decision tree or SVM
  • the processing system 4 generates and stores a mathematical model adapted to receive as input data the values of the cell parameters P r T * ' of any one cell of any one patient and generate in output a corresponding indication of the degree of aggressiveness of the tumoral mass present in this cell.
  • the processing system 4 has stored the first, the second and the third classifier, in addition to the list of the significant parameters, the significant regional parameters and the significant cell parameters.
  • the CAD system 1 is therefore ready to be used on an unknown patient, or on a subject for which there is no information on the possible presence of tumoral masses.
  • the operations in blocks 10, 20 and 30 are performed, indicated here by 510, 520 and 530 respectively.
  • the processing system 4 determines (block 540) the values of the significant parameters (the latter having been selected during the operations in block 60 of figure 2 ) of the unknown patient.
  • the processing system 4 applies the first classifier to the values of the significant parameters of the unknown patient, generating (block 545) the three-dimensional likelihood map relative to the unknown patient.
  • the processing system 4 then performs the operations for the blocks 80, 90, 100, 110 and 120 (now indicated overall by 550), so as to determine the three-dimensional regions of voxels of the unknown patient.
  • the processing system 4 determines (block 560) the values of the significant regional parameters (namely, the regional parameters selected in block 150 of figure 2 ) relative to the unknown patient.
  • the processing system 4 then applies the second classifier to the values of the significant regional parameters of the unknown patient, generating (block 570) an aggressiveness map relative to the unknown patient, which associates, with each three-dimensional region of voxels, a corresponding indication of the degree of aggressiveness.
  • the processing system 4 carries out the operations in blocks 171-172 (now indicated overall by 580), so as to determine the cells of the relevant three-dimensional regions of the unknown patient.
  • the processing system 4 considers relevant the three-dimensional regions of voxels having degrees of aggressiveness higher, for example, than 3+3.
  • the processing system 4 determines (block 590) the values of the significant cell parameters (namely, the regional cell parameters selected in block 175 of figure 2 ) relative to the unknown patient.
  • the processing system 4 then applies the third classifier to the values of the significant cell parameters of the unknown patient, generating (block 600) a cell aggressiveness map relative to the unknown patient, which associates, with each cell of each relevant three-dimensional region of the unknown patient, a corresponding indication of the degree of aggressiveness.
  • the present CAD system allows identification and characterization of a prostate tumoral mass without the direct intervention of an expert.
  • the present CAD system is able to select, without requiring the intervention of an expert, the three-dimensional regions having a high probability of including tumoral masses, and characterize their aggressiveness, on the basis of the corresponding values of the significant regional parameters.
  • the structural parameters and the regional parameters can be calculated also on small-dimension regions (for example, formed of only a few voxels); in particular, with reference to the regional parameters, this allows very precise indications to be provided on the aggressiveness and heterogeneity of a tumour, assisting the radiologist in the choice of the treatment most suited to the patient in question.
  • the division of the relevant three-dimensional regions into cells allows the precision level of the analysis to be further increased.
  • said CAD system can be adapted to be used in body regions different from the prostate, such as, for example, breast, rectum and lungs.
  • processing system 4 may be configured to generate further quantities, with respect to the preceding description. For example, the processing system 4 may generate, for each three-dimensional region of voxels of the unknown patient, a corresponding PIRADS (Prostate Imaging Reporting and Data System) score.
  • PIRADS Prostate Imaging Reporting and Data System
  • the operations carried out by the present processing system 4 may differ from what is described.
  • the three-dimensional regions may be determined differently from what is described; for example, the processing system 4 may look for the absolute maximum and/or one or more relative maximums of each three-dimensional likelihood map and select the relative neighbourhoods, independently of the fact that these maximums exhibit probability values higher than a threshold.
  • each three-dimensional region it is possible, for example, for each three-dimensional region to be formed of the voxels that are less than a predefined distance from a corresponding voxel that exhibits an absolute or relative maximum of probability.
  • the step of determination of the three-dimensional regions includes an additional step with respect to what is described, in which a three-dimensional region is discarded if it does not meet a further criterion, for example relative to the form (for example, concavity/convexity) .

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Physics & Mathematics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Pathology (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Data Mining & Analysis (AREA)
  • Radiology & Medical Imaging (AREA)
  • Primary Health Care (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Biophysics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Veterinary Medicine (AREA)
  • Epidemiology (AREA)
  • Evolutionary Computation (AREA)
  • Probability & Statistics with Applications (AREA)
  • Software Systems (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Databases & Information Systems (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)
EP19730217.7A 2018-05-08 2019-05-08 Tumoral mass detection system based on magnetic resonance imaging Active EP3791361B1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
IT102018000005163A IT201800005163A1 (it) 2018-05-08 2018-05-08 Sistema di rilevamento di masse tumorali basato sulla risonanza magnetica per immagini
PCT/IB2019/053783 WO2019215637A1 (en) 2018-05-08 2019-05-08 Tumoral mass detection system based on magnetic resonance imaging

Publications (2)

Publication Number Publication Date
EP3791361A1 EP3791361A1 (en) 2021-03-17
EP3791361B1 true EP3791361B1 (en) 2022-08-03

Family

ID=63244730

Family Applications (1)

Application Number Title Priority Date Filing Date
EP19730217.7A Active EP3791361B1 (en) 2018-05-08 2019-05-08 Tumoral mass detection system based on magnetic resonance imaging

Country Status (7)

Country Link
US (1) US11983884B2 (pt)
EP (1) EP3791361B1 (pt)
ES (1) ES2929051T3 (pt)
IT (1) IT201800005163A1 (pt)
LT (1) LT3791361T (pt)
PT (1) PT3791361T (pt)
WO (1) WO2019215637A1 (pt)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102600401B1 (ko) * 2021-06-03 2023-11-10 주식회사 크레스콤 분류 및 분할을 이용한 의료 영상 분석 방법, 장치 및 컴퓨터 프로그램

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080125643A1 (en) * 2006-11-24 2008-05-29 Qview, Inc. Processing and displaying dynamic contrast-enhanced magnetic resonance imaging information
WO2013086026A1 (en) * 2011-12-05 2013-06-13 The Johns Hopkins University System and method of automatically detecting tissue abnormalities
WO2014174317A2 (en) * 2013-04-26 2014-10-30 St George's Hospital Medical School Processing imaging data to obtain tissue type information
EP3081955A1 (en) * 2015-04-13 2016-10-19 Commissariat A L'energie Atomique Et Aux Energies Alternatives Mri method for determining signature indices of an observed tissue from signal patterns obtained by motion-probing pulsed gradient mri
US20190183429A1 (en) * 2016-03-24 2019-06-20 The Regents Of The University Of California Deep-learning-based cancer classification using a hierarchical classification framework

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2018514844A (ja) * 2015-03-11 2018-06-07 シーメンス アクチエンゲゼルシヤフトＳｉｅｍｅｎｓ Ａｋｔｉｅｎｇｅｓｅｌｌｓｃｈａｆｔ 細胞画像および映像の深層畳み込みネットワークベースの分類のためのシステムおよび方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080125643A1 (en) * 2006-11-24 2008-05-29 Qview, Inc. Processing and displaying dynamic contrast-enhanced magnetic resonance imaging information
WO2013086026A1 (en) * 2011-12-05 2013-06-13 The Johns Hopkins University System and method of automatically detecting tissue abnormalities
WO2014174317A2 (en) * 2013-04-26 2014-10-30 St George's Hospital Medical School Processing imaging data to obtain tissue type information
EP3081955A1 (en) * 2015-04-13 2016-10-19 Commissariat A L'energie Atomique Et Aux Energies Alternatives Mri method for determining signature indices of an observed tissue from signal patterns obtained by motion-probing pulsed gradient mri
US20190183429A1 (en) * 2016-03-24 2019-06-20 The Regents Of The University Of California Deep-learning-based cancer classification using a hierarchical classification framework

Also Published As

Publication number Publication date
ES2929051T3 (es) 2022-11-24
WO2019215637A8 (en) 2020-02-27
LT3791361T (lt) 2022-11-25
IT201800005163A1 (it) 2019-11-08
EP3791361A1 (en) 2021-03-17
US20210219861A1 (en) 2021-07-22
WO2019215637A1 (en) 2019-11-14
PT3791361T (pt) 2022-10-14
US11983884B2 (en) 2024-05-14

Similar Documents

Publication Publication Date Title
Sun et al. Multiparametric MRI and radiomics in prostate cancer: a review
EP3367331A1 (en) Deep convolutional encoder-decoder for prostate cancer detection and classification
Huang et al. Analysis of tumor vascularity using three-dimensional power Doppler ultrasound images
US20150003706A1 (en) Probability mapping for visualisation and analysis of biomedical images
EP2987114B1 (en) Method and system for determining a phenotype of a neoplasm in a human or animal body
KR101711228B1 (ko) 신체, 특히, 종양의 용적 변화를 수반하는 병리 진행을 정량화하기 위한 방법
CN112508884B (zh) 一种癌变区域综合检测装置及方法
CN110785123B (zh) 采用改进的数据处理技术对组织异常进行体素内不相干运动mri三维定量检测
CN117530659B (zh) 基于光学相干层析成像技术的早期宫颈癌变辅助诊断工具
EP3791361B1 (en) Tumoral mass detection system based on magnetic resonance imaging
Tai et al. Cancer-net BCa: Breast cancer pathologic complete response prediction using volumetric deep radiomic features from synthetic correlated diffusion imaging
CN113160256A (zh) 一种多任务生成对抗模型的mr影像胎盘分割方法
Rajive Gandhi et al. A contrast adaptive histogram equalization with neural learning quantization (CAHE-NLQ) for blood clot detection in brain
Yang et al. Value of a combined magnetic resonance imaging-based radiomics-clinical model for predicting extracapsular extension in prostate cancer: a preliminary study
KOWSHIKA et al. Advanced logistic regression for detecting the brain tumour cells
Oghli et al. Automatic fetal biometry evaluation in ultrasound images using a deep learning-based approach
CN118229684B (zh) 一种鉴别肾上腺嗜铬细胞瘤的方法及鉴别系统
Chaphekarande et al. Machine learning based brain mri estimation method
US20230196563A1 (en) Magnetic resonance image processing method
SATHYAKALA et al. Unsupervised Classification for predicting Malignant Tumor cells in Brain using FCM Method
Lakmi et al. Identification of Brain Tumor and Extracting its’ Features through Processing of MRI
Fernandes et al. Prostate cancer risk assessment using a radiogenomic analysis
Khazendar Computer-aided diagnosis of gynaecological abnormality using B-mode ultrasound images
Kremer MRI Radiomics for Imaging Genomics, Risk Classification, and Prediction of Disease Progression in Autosomal Dominant Polycystic Kidney Disease
Arunachalam et al. Detection of Singularity in the Cell Nucleus of Synovial Sarcoma Using Wavelet Leaders

Legal Events

Date Code Title Description
STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: UNKNOWN

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE

PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE

17P Request for examination filed

Effective date: 20201119

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

AX Request for extension of the european patent

Extension state: BA ME

REG Reference to a national code

Ref country code: HK

Ref legal event code: DE

Ref document number: 40050709

Country of ref document: HK

GRAP Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOSNIGR1

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: GRANT OF PATENT IS INTENDED

INTG Intention to grant announced

Effective date: 20220301

GRAS Grant fee paid

Free format text: ORIGINAL CODE: EPIDOSNIGR3

GRAA (expected) grant

Free format text: ORIGINAL CODE: 0009210

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE PATENT HAS BEEN GRANTED

AK Designated contracting states

Kind code of ref document: B1

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

REG Reference to a national code

Ref country code: AT

Ref legal event code: REF

Ref document number: 1509361

Country of ref document: AT

Kind code of ref document: T

Effective date: 20220815

Ref country code: CH

Ref legal event code: EP

REG Reference to a national code

Ref country code: DE

Ref legal event code: R096

Ref document number: 602019017792

Country of ref document: DE

REG Reference to a national code

Ref country code: IE

Ref legal event code: FG4D

REG Reference to a national code

Ref country code: PT

Ref legal event code: SC4A

Ref document number: 3791361

Country of ref document: PT

Date of ref document: 20221014

Kind code of ref document: T

Free format text: AVAILABILITY OF NATIONAL TRANSLATION

Effective date: 20221007

REG Reference to a national code

Ref country code: SE

Ref legal event code: TRGR

REG Reference to a national code

Ref country code: NL

Ref legal event code: FP

REG Reference to a national code

Ref country code: ES

Ref legal event code: FG2A

Ref document number: 2929051

Country of ref document: ES

Kind code of ref document: T3

Effective date: 20221124

REG Reference to a national code

Ref country code: NO

Ref legal event code: T2

Effective date: 20220803

REG Reference to a national code

Ref country code: GR

Ref legal event code: EP

Ref document number: 20220402115

Country of ref document: GR

Effective date: 20221212

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: RS

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

Ref country code: LV

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

Ref country code: FI

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: PL

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

Ref country code: IS

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20221203

Ref country code: HR

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: SM

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

Ref country code: RO

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

Ref country code: DK

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

Ref country code: CZ

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

VS25 Lapsed in a validation state [announced via postgrant information from nat. office to epo]

Ref country code: MD

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

REG Reference to a national code

Ref country code: DE

Ref legal event code: R097

Ref document number: 602019017792

Country of ref document: DE

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: SK

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

Ref country code: EE

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

PLBE No opposition filed within time limit

Free format text: ORIGINAL CODE: 0009261

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: AL

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

26N No opposition filed

Effective date: 20230504

P01 Opt-out of the competence of the unified patent court (upc) registered

Effective date: 20230602

REG Reference to a national code

Ref country code: AT

Ref legal event code: UEP

Ref document number: 1509361

Country of ref document: AT

Kind code of ref document: T

Effective date: 20220803

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: SI

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: MC

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

REG Reference to a national code

Ref country code: BE

Ref legal event code: MM

Effective date: 20230531

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: MC

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20220803

Ref country code: LU

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20230508

REG Reference to a national code

Ref country code: IE

Ref legal event code: MM4A

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: IE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20230508

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: IE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20230508

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: BE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20230531

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: NL

Payment date: 20240526

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: GB

Payment date: 20240527

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: LT

Payment date: 20240418

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: DE

Payment date: 20240530

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: GR

Payment date: 20240530

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: CH

Payment date: 20240602

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: ES

Payment date: 20240603

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: AT

Payment date: 20240419

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: IT

Payment date: 20240423

Year of fee payment: 6

Ref country code: FR

Payment date: 20240527

Year of fee payment: 6

Ref country code: NO

Payment date: 20240530

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: PT

Payment date: 20240422

Year of fee payment: 6

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: SE

Payment date: 20240527

Year of fee payment: 6

Ref country code: MT

Payment date: 20240424

Year of fee payment: 6